Short answer
If you liked how Fooocus felt, HEISS UI brings that feel to today’s models. Pick Flux.2, Krea 2, Qwen-Image or Z-Image, write a prompt, and the settings already fit. It suggests a first model sized for your computer, and adds video, a phone studio and one searchable gallery.
Fooocus is still very good for SDXL, with its own inpainting, styles and Vary buttons, and it runs from 4 GB on NVIDIA. It won’t get newer models: its README says there are no plans to move past SDXL.
Pick HEISS UI if
- You want newer models: Flux.2, Krea 2, Qwen-Image, Z-Image, Chroma, HiDream and more.
- You’d like the same “just write a prompt” feel, with settings that fit each model.
- You want a first model picked to fit your graphics card or Mac.
- You want video too: Wan 2.2, HunyuanVideo 1.5 and MiniMax H3, in beta.
- You want to prompt from your phone and keep one gallery you can search.
Pick Fooocus if
- You work with SDXL and its fine-tunes and are happy there.
- You use its inpaint and outpaint, which run on its own inpaint model.
- You like its styles, Vary (Subtle and Strong) and Image Prompt.
- You have a 4 GB NVIDIA card.
Only in HEISS UI
- Today’s models. Flux.2, Krea 2, Qwen-Image, Z-Image and more, 21 image families in all.
- Video, in beta. Wan 2.2, HunyuanVideo 1.5 and MiniMax H3.
- A first model for your computer. The size that fits your graphics card or Mac, marked.
- Editing by instruction. Flux.2 and Qwen-Image 2.1 change an image from reference images and a sentence.
- A phone studio. Made for one thumb, with its own sign-in.
- Hidden. An encrypted place for private images.
At a glance.
| Feature | HEISS UI | Fooocus |
|---|---|---|
| Newest models | 21 image familiesFlux.2, Krea 2, Qwen-Image, Z-Image and more | SDXL onlyand its fine-tunes |
| Settings | Fit each modelset when you pick it | Tuned for SDXLwith quality and speed presets |
| First model | Picked for your computerthree sizes, the one that fits marked | SDXL, downloaded on first start |
| Video | Beta7 families | No |
| Upscale | One clickwith a compare slider | 1.5× and 2× |
| On a phone | Phone studiomade for one thumb | Browser page |
| Private images | Hiddenencrypted | No |
| Status | Active0.13.0 on 28 Sep 2026 | Long-term supportbug fixes only |
| Engine | Your ComfyUI | Built in |
| Image prompts | Reference imagesFlux.2 and Qwen-Image 2.1; start image for the rest | Image Prompt, Vary |
| Styles | No | Yes |
| Inpaint, outpaint | No | Yesits own inpaint model |
| Smallest GPU | 6 GB for an SDXL Lightning build8 GB for SDXL and the compact Flux.2 Klein | 4 GB NVIDIA |
| Licence | MIT | GPL-3.0 |
Checked 29 September 2026 against the Fooocus README and releases.
Where Fooocus stands.
Fooocus is lllyasviel’s prompt-first app, built to feel like the big online image generators, but offline: short prompts, good results, very few settings. Its README now reads “Limited Long-Term Support (LTS) with Bug Fixes Only”. It’s built entirely on SDXL, and the README says there are no current plans to add newer architectures, pointing people who want Flux to Forge, ComfyUI or SwarmUI. The last release, 2.5.5, is from August 2024.
Some forks carry it further. RuinedFooocus, for example, lists Flux, GGUF Flux and SD3 in its README. They’re separate projects with their own pace.
People asked for GGUF support to run Fooocus on smaller cards. The project stayed on SDXL.
The README warns about fake Fooocus websites and names the GitHub page as the only official source.
The same loop, newer models.
What made Fooocus good was that you didn’t have to know the sampler, the steps or the CFG. HEISS UI works the same way, per model. Pick Krea 2 Turbo and it runs at 8 steps with CFG 1. Pick Flux.2 Klein and it runs at 4. Pick an SDXL fine-tune and it gets 25 steps and CFG 7, or few-step settings when the file is a Lightning or DMD2 build. The numbers come from the official ComfyUI templates and model cards, and only the controls a model uses show up.
Around that loop: previews stream in while the image renders, ↑ in an empty prompt brings back recent prompts, / searches the gallery, and LoRAs stack per family with their trigger words. Surprise me and Vary this help when you’re out of ideas.
Where to start.
With no model installed, HEISS UI offers Krea 2, Flux.2 and SDXL, each in three sizes, and marks the one that fits your graphics card. One tap gets the model with its text encoder and VAE. Your Fooocus SDXL checkpoints work too: put them in ComfyUI’s models/checkpoints folder, or add the Fooocus folder to extra_model_paths.yaml.
How much memory each needs: 8 GB cards, 12 GB cards, and the model guides for Krea 2, Flux.2 Klein and Z-Image.
Where Fooocus is the better pick.
Fooocus’s inpainting uses its own model and is still one of the best for SDXL. Its styles and its Vary and Image Prompt buttons have no direct match here. HEISS UI has a start image with a Change slider for every image model, and reference images for the edit models Flux.2 and Qwen-Image 2.1, which can do a lot of what Vary did with a sentence of instruction.
Questions.
Is Fooocus still maintained?
Only with bug fixes. Its README calls it limited long-term support, built on SDXL, with no current plans for newer architectures. The last release is 2.5.5 from August 2024.
Can Fooocus run Flux or Z-Image?
Not the original. Forks such as RuinedFooocus list Flux support. For current models with a similar prompt-first feel, HEISS UI runs them on ComfyUI.
Can I use my Fooocus models in HEISS UI?
Yes. SDXL checkpoints work as they are. Put them in ComfyUI’s models/checkpoints folder, or add the Fooocus models folder to ComfyUI’s extra_model_paths.yaml.
Does HEISS UI have styles like Fooocus?
No. It has recent and starred prompts, Surprise me and Vary this, and each model’s own settings, but no style presets.
How much graphics memory do I need?
A card with 8 GB runs SDXL and the compact Flux.2 Klein, 12 to 16 GB is comfortable, and 24 GB or more runs nearly everything. On a Mac, 16 GB runs SDXL.
Sources: Fooocus README, Fooocus releases, issue #3809, RuinedFooocus, HEISS UI README.
